Innovation in Lending

After Six Million Credit Applications, The Big Thing We Learned: AI Makes Lenders More Money

Zest AI team

August 30, 2021

Thanks to our customers, Zest AI has put more AI-driven credit-underwriting models into production than anyone in the industry. As a result, Zest-built models now score close to $125 billion in lending portfolios across all credit products - personal loans, credit cards, auto loans, student loans, and mortgages - and we’ve generated more than six million scores to date. So we learned a few things (some the hard way) about AI-driven lending.


Download the Zest Guide: Six Million Credit Apps Later - What Have We Learned About AI-Driven Lending

The biggest lesson of all: Machine learning works. It is simply better at predicting borrower defaults better than legacy models like FICO. Lenders who have switched to AI see their predictive accuracy go up 10% to 20% using more data and better machine learning math. Moreover, the improvements run across all credit bands and customer profiles. See the chart below. The bottom axis is AUC, (short for "area under the curve") a standard measure of statistical accuracy in a classifier model. A higher number means the model is better able to rank a random positive instance higher than a random negative instance. This translates into more and better loans and faster decisions.

We’ve learned two more things: ML models, rather than perpetuating the racial bias of traditional credit scores and human underwriters, can be optimized for greater fairness. We’ve written about this before. And, machine learning-based models can consistently meet or exceed the standards imposed by federal and state banking regulators.

Results like these will force a much-needed change in the status quo in credit scoring. The U.S. suffers from broad and deep gaps in access to affordable credit. A disproportionate number of  Black and Latinx applicants are among the 40% of the country who are hard to score, credit invisible, or have artificially depressed credit scores because of how the system works today.

Lenders who step past their doubts can capture the most significant competitive advantages in pricing and efficiency -- and improve social outcomes more widely. The top 20 banks and top 5 credit unions may have the money and people to build and run their AI-driven lending. But the other 14,000 financial institutions in the United States will need help. Our team has helped dozens of lenders make the transition to AI-driven lending. We’ve packed some of what we’ve learned into the latest Zest Guide, “Six Million Credit Applications Later, What Have We Learned About AI-Driven Lending?”

Topics covered include:

  • Why financial inclusion is sound business
  • The advantage of trended data
  • Why counterfactual portfolios are key
  • How to improve fair lending outcomes
  • Why it’s crucial to do proactive monitoring of credit scoring
  • Insights on change management and innovation

Download the Zest Guide: Six Million Credit Applications Later - What We’ve Learned About AI-Driven Lending

----

Read more buyer's guides on our Resources page.

Thank you for subscribing!
Something went wrong while submitting the form.